System Configuration

[Tony Hammitt]

I was talking about using Python for the configuration parameter files
themselves.  Like /etc/xinetd.conf etc. would become:

#!/usr/bin/python2

from xinetd import *

defaults = {
        'instances'		:60
        'log_type'		:"SYSLOG authpriv"
        'log_on_success'	:"HOST PID"
        'log_on_failure'	:"HOST"
}

This would import everything from the xinetd directory (once it, too is
set up as Python) and assign to the dictionary called defaults the named
entries that were in the original config file.  Then you can use the
setup directly from C by embedding calls out to python in the C code
that just execute the configuration file and make the top-level objects
available to the calling program.

Writing the manipulation tools in python would be good, too since it can
test the syntax and reload modules as needed.

So, anyway, this is still a proposal for a truly massive amount of work,
so it may be good to just have the basic distro in place first =-]

Later,
	Tony

Brian Densmore wrote:
> 
> Yes, I like the idea of using Python or some other scripting language to
> create configuration tools. Especially if the language is extensible,
> like Python. Python is nice because there are a lot of drop in
> capabilities to add GUI interfaces very easily. I really don't know what
> state XML is in.
> 
> I got my "new" CD burner today so I'll start playing with burning a
> rough draft of a distro. Based on the LFS and/or gentoo systems. I
> picked up a refurb'ed Plextor, from Hyper Microsystems
> (http://www.hypermicro.com/). They've been very good so far. Not the
> cheapest place, but good service and quality products.
> LFS is real bare bones. I need to pull down several add-ons for it.
> 
> I'll do the installer in Python and see how it works. Can I use the
> KCLUG logo for the lilo logo screen?
> 
> Brian
> 
> > -----Original Message-----
> > From: Tony Hammitt [mailto:tony at speedscript.com]
> > Sent: Friday, February 08, 2002 11:52 AM
> > Cc: kclug at kclug.org
> > Subject: Re: System Configuration [Was: Changing IP Addresses]
> >
> >
> > I did some further thinking about this and it seems to me
> > that we could
> > just have Python code as the configuration file format.  It
> > has variable
> > assignments, heirarchy, class structure and system call
> > capabilities and
> > Python is installed by default on most distros.
> >
> > The only issue is that the kit is like 22MB on disk, which is
> > a lot for
> > a handheld =-]  But I'm sure that there's some subset of the codebase
> > that could be pruned down to run just the configuration file things.
> >
> > Python's tight integration with C would make extending the code pretty
> > simple compared to other approaches like XML where we'd have
> > to write the
> > integration code ourselves.  That and it can have normal
> > looking comments.
> >
> > Anyway, it's just a thought.  Comments?  (Mike?)
> >
> > Later,
> >       Tony
> >
> >
> > Brian Densmore wrote:
> > >
> > > Tony,
> > >    I'd like to discuss this. There may be a way to incorporate this
> > > into the new distro.
> > >
> > > Brian
> > >
> > > PS. I won't be able to make tonight's meeting. The power
> > outages have
> > > caused
> > > some rescheduling of my activities. :'(
> > >
> > > I almost have the linuxfromscratch installed. Had a minor problem
> > > building
> > > the first compiler (somehow I did build a good makefile the
> > 1st time).
> > > Only
> > > 35 packages to go!
> > >
> > > > -----Original Message-----
> > > > From: Tony Hammitt [mailto:thammitt at kc.rr.com]
> > > > Sent: Tuesday, February 05, 2002 6:14 PM
> > > > Cc: kclug at kclug.org
> > > > Subject: System Configuration [Was: Changing IP Addresses]
> > > >
> > > >
> > > > OK, so what it comes down to is that we need to start using
> > > > something like XML with validators for all of the system and
> > > > application configuration files.  We need a universal standard
> > > > way of storing and updating information.  Unfortunately, this
> > > > would be an incredible, major effort to get started.  Also, XML
> > > > itself isn't really perfect for the task since commenting is a
> > > > pain and the validation engine would need updated to be useful
> > > > when there is no network.
> > > >
> > > > But it would allow us to define a set of acceptable values for
> > > > the configuration parameters, validate that they make sense as
> > > > a set and provide an easy way to add functionality without
> > > > breaking the old config files or programs.
> > > >
> > > > So, maybe it would take a few years to get it all working, but
> > > > then those of us who like hand-editing files can do so and
> > > > those GUI config tool fans could have a consistent interface.
> > > > What I like best about the idea is that the config files would
> > > > all be in the same format, so people wouldn't have to learn to
> > > > read each file format, like they do now.
> > > >
> > > > Unfortunately, the whole scheme is going to seem too much like
> > > > the awful windoze registry for some people to accept.  (As if
> > > > _everything_ about M$ is terrible, not just 99.97% of it :).
> > > >
> > > > That's what I'd do about the problem.  Heck, that's what I DID
> > > > about the problem for my code.  I have a configuration file
> > > > library I use for several projects.  It's LGPL if anyone is
> > > > interested, I still have to write the validator and some other
> > > > features, but it won't be hard.  Let me know...
> > > >
> > > > Regards,
> > > >
> > > >       Tony
> > > >
> > > > Jonathan Hutchins wrote:
> > > > >
> > > > > ----- Original Message -----
> > > > > From: "Joshua Bergland" <kclug at mrj412.com>
> > > > >
> > > > > > I just don't buy the argument that making Linux user
> > friendly will
> > > > > > hinder the OS.
> > > > >
> > > > > I wouldn't buy or make that argument either.  However,
> > > > "user friendly" and
> > > > > "Microsoft Copycat" are not equivalent terms.  GUI tools
> > > > are nice, and
> > > > > they're a good step toward making Linux more accessible to
> > > > the average
> > > > > user - especially average users who have never known an
> > > > environment prior to
> > > > > Windows95 where configuration by command line was assumed.
> > > > >
> > > > > However, those tools should work _with_ standard
> > > > configuration files, not
> > > > > strike out on their own and overwrite configurations from
> > > > the standard text
> > > > > files - as linuxconf definitely does.  They should work in
> > > > a way that helps
> > > > > the user see what's going on in the configuration, and
> > > > possibly offer direct
> > > > > access to the configuration files themselves - but not
> > > > without showing
> > > > > exactly where that information is being stored.  (This is
> > > > one of the major
> > > > > pains in the Microsoft world - you can't ever be sure a
> > program is
> > > > > completely uninstalled, because it leaves fewmets all over
> > > > the operating
> > > > > system.  Likely as not, a re-install will miraculously
> > > > recover the settings
> > > > > you made in the original, even if those settings cause the
> > > > program to
> > > > > crash.)
> > > > >
> > > > > > Unless things change, I can't forsee Linux ever
> > > > pentrating the desktop
> > > > > > market. Of course, this may not be a goal of Linux either...
> > > > >
> > > > > Since Linux, in and of itself, is not a market
> > competitor but more a
> > > > > philosophy and a development model, I find debate about
> > > > "market share"
> > > > > pretty obtuse.  I know what you mean, but I'm not overly
> > > > worried about "our
> > > > > side winning".
> > > > >
> > > > majordomo at kclug.org
> > > >
> > > >
> > > > majordomo at kclug.org
> > > >
> > >
> > majordomo at kclug.org
> >
> >
> > majordomo at kclug.org
> >
>